1. Field of the Invention
The present invention relates to a storage, a storage system having a data fetching method, particularly a trace function, to trace a command sequence in an input/output (I/O) interface, and a method of fetching trace information thereof.
2. Description of the Related Art
In the storage system, information obtained by tracing a command sequence in an I/O interface is useful for failure analysis and performance analysis. In the failure analysis, the data is used as follows. A command sequence before and after occurrence of a failure is used to detect a process to the failure and direct causes of the failure. Additionally, the data is used in the performance analysis as follows. A pattern of the command sequence issuance is determined to detect processing in the storage system which exerts influence on the performance.
In a storage system connected to a main frame computer, trace information of such a command sequence is in general fetched by, for example, a utility “generalized trace facility (GTF)” of IBM. Moreover, there exists a method to fetch the trace information using a device called “SCSI analyzer”. The SCSI analyzer can be connected to a storage system to fetch trace information.
However, GTF is a function installed only in a main frame system, and an equivalent trace function is not satisfactory supported in an open system at present. Therefore, there possibly occurs a case in which a failure taking place in an open system cannot be easily analyzed. Moreover, when the SCSI analyzer is used, it is necessary to once stop a path on the host side connected to the pertinent SCSI path. This possibly exerts adverse influence on business of the customer, and hence it is difficult to fetch trace information at an appropriate time.
To solve the problem of the SCSI analyzer, for example, JP-A-2000-187621 describes a technique in which information on a bus of the SCSI and information on a processor bus are traced and the trace is stopped at occurrence of an abnormality. However, if a trace area reserved is insufficient, new trace information overlaps old trace information or the trace is stopped when the trace area becomes full of information. Consequently, expected information cannot be fetched depending on cases.
To solve the problem of the trace area, JP-A-11-212821 describes a technique in which the trace information is fetched in a trace buffer with a limited capacity by using a data compression/expansion function of a magnetic tape controller. However, also in this technique, since the trace information is fetched in the buffer with a physically limited capacity, even when trace information is fetched for a long period of time for performance analysis, there occurs a case in which information is insufficient. This disadvantageously leads to a problem that the performance analysis cannot be conducted.
Furthermore, in a case of a storage system including a plurality of processors to process data I/O operations between hosts and storages, there occurs a problem that even trace information of the respective processors is fetched, a relationship of a sequence of trace information cannot be easily recognized.
In short, the problems will be listed in order as follows.    (1) Establishing a unit to fetch long-term trace information.    (2) Establishing a unit to fetch trace information having a consistent relationship during execution of concurrent processing by a plurality of processors.    (3) Establishing a trace information fetching unit not exerting adverse influence on business of customers.
The inventor considers that a fundamental cause of three problems above is that all trace information in the storage system is treated as data dedicated to the storage system. Therefore, the inventor recognizes that a dedicated buffer and a dedicated fetching tool are necessary in the storage system and hence cause various restrictions.